Waking up to the news, NVIDIA’s Thor chip has been delayed yet again! Originally scheduled for mass production in mid-2024, the Thor chip has now been significantly postponed to “expected to launch mid-next year, and only the entry-level version.” This has left many automakers, especially companies like Xiaopeng and Li Auto, who were counting on the Thor chip for advanced intelligent driving, in a state of confusion.
I find this situation quite interesting; even a chip giant like NVIDIA can experience “delays.” To be honest, since the Thor chip was announced in 2022, it has been delayed countless times. NVIDIA initially boasted that this chip could achieve 2000 TOPS, which is eight times that of Orin, and could “replace six chips with one.” What happened? Now the computing power has shrunk to 700 TOPS, and the integrated design of “one chip replacing six” has become a fantasy, turning it into just an ordinary intelligent driving chip.
Automakers’ Plans Completely Disrupted
This has left automakers in a tough spot. Xiaopeng’s P7+ originally intended to use the Thor chip directly, but due to the delay, they had to switch to a dual Orin-X configuration. Xiaopeng’s product manager even complained on social media: “Thor has been delayed again, and there is still no confirmed SOP date. At this rate, seeing it in 2026 would be good enough.”
Li Auto is in the same boat; they originally planned to deploy their 4 billion parameter VLA model using the Thor chip, but the delay has pushed back the entire L series model refresh. I heard that many of Li Auto’s suppliers were notified last year that the upgraded L series, originally set to launch in March this year, has been collectively postponed due to the Thor chip not being delivered on time.
BYD and Zeekr also had intentions to collaborate with the Thor chip, but now they have mostly changed their plans. After all, 2025 is being called the “Year of Equal Rights in Intelligent Driving,” and intelligent driving features are gradually becoming standard in mid-to-high-end models; automakers cannot afford to wait for NVIDIA’s slow progress.
Domestic Chips Seize Opportunities
However, NVIDIA’s delays have provided opportunities for domestic chip companies. Originally, NVIDIA’s Orin series held nearly 40% of the installed market share, but the market landscape is shifting from “one dominant player” to “diverse competition.”
Horizon Robotics has seized the opportunity, with their Journey 6P (J6P) achieving 560 TOPS, comparable to the dual Orin-X solution’s 508 TOPS. They have already established partnerships with leading automakers like BYD, Li Auto, Volkswagen, and GAC. Particularly, the investment by NavInfo in JianZhi Robotics is aimed at leveraging JianZhi’s full-stack mass production capabilities on the Horizon J6E/M platform.
Aixinyuan is also racing to push products, with their ADAS chip M55H already in mass production, M57 receiving mass production approval, and M76H entering the pre-mass production stage. The next generation of high-end chips will support advanced technologies like end-to-end and VLM visual large models, appearing quite competitive.
Mobileye has also been active, with the EyeQ6 Lite chip already being delivered, and it is expected to reach millions of units in the coming years.
Automakers Begin Developing Their Own Chips
Interestingly, “Weilai, Xiaopeng, and Li Auto” have all started developing their own chips. Xiaopeng’s “Turing” chip successfully completed tape-out in August this year and is currently undergoing testing and validation, with full-stack NGP already running on the chip. NIO’s “Shenji NX9031” has also successfully completed tape-out and is a vehicle-grade 5nm intelligent driving chip set to be installed in the ET9. Li Auto’s “Schumacher” chip is also under development, designed with a 5nm process and manufactured by TSMC.
Why are automakers developing their own chips? Li Bin stated quite frankly that they spent a lot of money on NVIDIA chips last year, and developing their own chip can replace four chips, significantly reducing costs. Li Xiang from Li Auto has also mentioned that if they develop their own inference chip, they can achieve costs similar to Tesla’s, as they control the algorithms.
However, the investment in developing their own chips is indeed substantial; the R&D cost for a 7nm process, high-performance SoC with over 100 TOPS exceeds $100 million. Liu Yudong from Chentao Capital has noted that if the shipment volume of self-developed chips is below 1 million units, it is challenging to achieve a balance between investment and return. Therefore, automakers need to calculate their costs carefully.
Industry Competition Landscape is Changing
I believe that this delay of the Thor chip, while appearing to be a typical failure on NVIDIA’s part, has actually become a catalyst for the autonomy of China’s intelligent driving chip industry. Automakers are waking up from their dependence on NVIDIA’s technology and are beginning to embrace diverse supply chains and self-research.
Now, the competition in intelligent driving is no longer merely about chasing computing power parameters and hardware arms races; it is shifting towards actual scenario experiences, data iteration efficiency, and the synergy of software and hardware capabilities. The Chinese industrial chain, based on a deep understanding of local complex road conditions and rapid algorithm iteration capabilities, is establishing a unique competitive advantage.
For instance, Li Auto’s “VLA Driver Large Model” launched in August and Xiaopeng’s new P7, which features a new generation VLA architecture with latency under 100ms, are advantages for domestic intelligent driving companies. Huawei and NIO are also working on world models, using cloud-simulated data to help large models understand the world.
In summary, the delay of the Thor chip has made automakers realize that they cannot put all their eggs in one basket. Diversifying the supply chain, strengthening self-research, and building a more resilient industrial ecosystem is the long-term strategy.
What do you think about this? Feel free to share and discuss, and follow me for daily automotive news updates.